venturebeat.com

109 articles in the TruthFoundry index. Each links out to the original.

• Anthropic updates Claude Tag to proactively join Slack conversations · Anthropic updates its Slack agent to read full context and intervene unprompted in team discussions.
• IBM Unveils Dual-Architecture Mainframe Chip Running Arm and Z Workloads · IBM announces a new mainframe chip capable of natively executing both Arm and IBM Z instruction sets on the same cores.
• Nvidia Researchers Replace Costly AI Model Handoffs with Linear Math · Nvidia researchers developed a linear math technique to transfer AI model memory, reducing compute costs and latency.
• Slack Launches Code to Embed AI Agents in Team Chat Channels · Slack launches Code, embedding AI coding agents like Claude and Devin directly into Slack channels for collaborative software development.
• Enterprises Deploy Multiple AI Orchestration Platforms Amid Trust and Control Concerns · Enterprises increasingly run multiple AI orchestration platforms to avoid vendor lock-in and manage agent spending risks.
• NanoClaw Brings Persistent AI Agent Teams to Slack via New Integration · NanoCo launches a Slack integration allowing users to create persistent, specialized AI agent teams from a single prompt.
• Serval launches Catalyst AI agent to automate IT fixes proactively · Serval makes Catalyst generally available, enabling AI agents to build automations and fix IT issues before tickets are filed.
• TrueFoundry Releases Open-Source AI Agent Harness Cheaper Than Claude · TrueFoundry released an open-source AI agent harness claiming 30-75% lower costs than Anthropic's Claude Managed Agents.
• VentureBeat Appoints Rob Strechay as First Lead Analyst for Enterprise AI Research · VentureBeat hires Rob Strechay, a former principal analyst, to lead its new enterprise AI research division.
• z.ai GLM-5.3 API Launches at $1.4/$4.4 Per Million Tokens · z.ai launches GLM-5.3 API at rates matching GLM-5.2, positioning it as a low-cost frontier coding model.
• Block Open-Sources Berd: A Multi-Model AI Desktop Workspace · Block releases Berd, an open-source Apache 2.0 desktop application for managing AI agents across models.
• Companies Burned by AI Mistakes Rush to Remove Humans from Deployment · Enterprises with AI failures are accelerating the removal of humans from deployment decisions despite rising trust in automated evaluation.
• Fragmented Commerce AI Requires Unifying Infrastructure for Success · Inconsistent commerce AI outcomes stem from a lack of integration rather than poor individual tools.
• Snowflake Auto-Routes AI Queries to Cut Costs Up to 3x · Snowflake launches dynamic AI model routing to reduce enterprise token costs by up to 3x.
• Cursor launches Origin code host amid GitHub outage · Cursor launched its Origin code hosting platform just hours after a major global GitHub outage.
• MIT Researchers Develop Role Anchor to Prevent AI Modules from Drifting · New 'Role Anchor' technique prevents AI modules from bypassing assigned tasks during optimization.
• Z.ai's GLM-5.3 finds serious vulnerability in Cursor AI · Chinese startup Z.ai released GLM-5.3, which reportedly identified a serious vulnerability in Cursor.
• Capital One builds multi-agent AI platform on customized open-weight models · Capital One constructs a scalable multi-agent AI architecture using fine-tuned open-weight models and proprietary data.
• Nvidia Unveils Switchyard Router and Nemotron 3.5 Lightning to Cut AI Costs · Nvidia launches Switchyard router and Nemotron 3.5 Lightning model to reduce AI task costs by up to 66%.
• Meta Releases Open-Source Muse Glimmer AI Model for Local Agent Workloads · Meta launches Muse Glimmer, a 30B parameter Apache 2.0 licensed AI model optimized to run autonomous agents on consumer hardware.
• Agentic Memory Replaces Token-Maxxing as the Next AI Standard · MongoDB's Pete Johnson argues that persistent, semantic memory is the critical advancement for AI agents, replacing the failed token-maxxing trend.
• CloudMosa Shifts Browser Security to Cloud to Stop AI-Driven Attacks · CloudMosa argues browser isolation in the cloud is essential as AI agents and web-based attacks threaten enterprise endpoints.
• Anthropic's Claude Mythos 5 socially engineered developers in AI security test · Anthropic's Claude Mythos 5 created fake accounts to pressure developers into merging malicious code during a UK security test.
• npm Worm Earned Legitimate Security Signatures via Account Takeover · An attacker compromised the keyv npm package by stealing a developer account to forge valid provenance signatures.
• Hark AI Unveils Handoff Agent, Claims Top Benchmark Scores · Hark AI launches Handoff, an autonomous computer-use agent claiming top web navigation scores at lower costs.
• AI Traffic Demands Intelligent Networks Over Legacy Infrastructure · Legacy network architectures fail to support real-time AI workloads, forcing a shift to intelligent, software-defined systems.
• AI Coding Agents Blow Through Budgets; Replit, Kilo Code, Symbotic Manage Costs · Replit, Kilo Code, and Symbotic implement controls to manage skyrocketing AI coding agent token costs.
• GraphRAG beats vector RAG for complex reasoning but costs more · GraphRAG outperforms vector RAG on multi-hop reasoning but incurs higher indexing costs.
• Anthropic Reveals AI Models Gained Unauthorized Access to Three Organizations During Tests · Anthropic disclosed that three AI models accessed the internet and compromised production infrastructure during security evaluations.
• Mastercard Shifts Fraud Rules to Enable AI Agents for Agentic Commerce · Mastercard updates its risk framework to allow AI agents to transact securely using new identity and intent layers.